Several studies have shown an association between the consumption of energy-dense foods, such as sweets and salty snacks, and excessive weight in pre-school children( Reference Durao, Severo and Oliveira 1 ). On the other hand, diets high in fruits and vegetables (F&V) have been associated with reduced risk for obesity in children( Reference Moschonis, Kalliora and Costarelli 2 ). The WHO and the US Department of Agriculture suggest the sparing consumption of unhealthy snacks, while the recommendation for F&V is at least five servings (approximately 400 g) daily( 3 ). However, intake seems to fall well short of these guidelines, both in the case of unhealthy snacks( Reference Nicklas and Johnson 4 ), as well as in the case of vegetables( Reference Patterson, Block and Rosenberger 5 ), with consumption among young children being particularly low( Reference Dennison, Rockwell and Baker 6 ). In line with these observations, the ToyBox-study revealed that European pre-school children’s intake exceeds the recommendation regarding unhealthy snacks, with consumption varying from 53·3 g/d in Greece to 73·1 g/d in Belgium( Reference De Craemer, Lateva and Iotova 7 ).
Improving children’s dietary habits as early as possible is particularly important. Eating behaviour is formed in childhood, tracks over childhood( Reference Kelder, Perry and Klepp 8 ) and persists into adulthood( Reference Lien, Lytle and Klepp 9 ). Thus, the adoption of healthy instead of unhealthy food choices early in life could provide lifelong benefits. In addition, there is growing evidence that poor diet in childhood can lead to health problems commonly observed in adults, such as diabetes( Reference Whincup, Gilg and Papacosta 10 ), obesity( Reference Klesges, Klesges and Eck 11 ) and CVD( Reference Moller, Taubert and Allen 12 ). As a result, interventions targeting early childhood, such as the Toy-Box study, could offer the maximum health benefits. Furthermore, consuming healthy snacks such as F&V instead of unhealthy snacks may contribute in controlling pre-schoolers’ energy intake and body weight, since energy-dense food consumption has been found to be associated with higher BMI in young children( Reference van Stralen, te Velde and van Nassau 13 ).
In the ToyBox-study, focus groups with parents and teachers of pre-school children were executed in six European countries. The findings of the focus groups identified parental modelling, availability of healthy snacks and certain parenting practices as determinants of pre-schoolers’ dietary habits (report submitted to the European Commission). Following the intervention mapping protocol( Reference Bartholomew, Parcel and Kok 14 ), these findings were translated into practical strategies and were considered in the design of the ToyBox-intervention( Reference Manios, Androutsos and Katsarou 15 ).
So far, analysis of the results of the ToyBox-intervention has shown significant improvements in pre-schoolers’ diet quality( Reference Pinket, De Craemer and Huybrechts 16 ). Because family dynamics including family rules, modelling support and encouragement are important determinants of children’s health behaviour( Reference Wilfley, Kass and Kolko 17 ), the present study aimed to shed light on the family-related mechanisms mediating the effect of the ToyBox-study on pre-school children’s consumption of healthy and unhealthy snacks.
Τhe ToyBox-study (www.toybox-study.eu) aimed to develop, implement and evaluate a kindergarten-based, family-involved intervention to prevent obesity in pre-school children in six European countries (Belgium, Bulgaria, Germany, Greece, Poland and Spain). Pre-school children and their families were recruited from 309 kindergartens from three socio-economic groups, following a standardized approach, as described elsewhere by Manios et al.( Reference Manios, Androutsos and Katsarou 15 ). The study was registered at clinical_trials.gov (identifier: NCT02116296). The study design has been described in detail elsewhere( Reference Manios, Androutsos and Katsarou 15 , Reference Manios, Grammatikaki and Androutsos 18 ).
The ToyBox-intervention had a cluster-randomized design and was conducted during the school year 2012–2013. Four energy balance-related behaviours were targeted, namely drinking, snacking, physical activity and sedentary behaviours, as well as their determinants( Reference van Stralen, te Velde and van Nassau 13 , Reference Manios, Androutsos and Katsarou 15 , Reference De Craemer, De Decker and De Bourdeaudhuij 19 ).
Four levels were used for implementation of the snacking component of the ToyBox-intervention. Level 1 included the availability of healthy snacks at class/kindergarten; level 2 included children’s daily consumption of a healthy snack; level 3 included the execution of interactive classroom activities using a kangaroo hand-puppet as a role model; and level 4 included the delivery of two newsletters, two tip-cards and one poster to the parents/caregivers via the teachers. These newsletters, tip-cards and poster included key messages and practical tips on healthy snacking (e.g. F&V) instead of unhealthy choices (e.g. sweets and/or salty snacks)( Reference Manios, Androutsos and Katsarou 15 ). Levels 1 and 2 were conducted for the whole duration of the school year 2012–2013, whereas levels 3 and 4 were conducted over six weeks (i.e. during the first focus period between weeks 9 and 12, and during the repetition period between weeks 21 and 22)( Reference Manios, Grammatikaki and Androutsos 18 ). The ToyBox-intervention was implemented by kindergarten teachers, who attended three training sessions by the research staff of minimum one hour per session( Reference Androutsos, Katsarou and Payr 20 ).
The pre-test measurements were conducted in May/June 2012 and the post-test measurements were conducted after one year (i.e. May/June 2013). Parents/caregivers signed a consent form to participate in the study and were asked to fill out the Primary Caregivers’ Questionnaire (PCQ) regarding sociodemographic factors, lifestyle behaviours and perinatal factors, as well as an FFQ( Reference Huybrechts, De Backer and De Bacquer 21 ).
Snack consumption was assessed by combining the frequency as well as the average consumption for each item. Determinants of pre-schoolers’ snack consumption were self-reported by the parents/caregivers in the PQC, using Likert-type questions. More details on the questions used for assessing snack consumption as well as the determinants can be found in the online supplementary material, Supplemental Tables 1 and 2, while the questionnaire is available online at the ToyBox-study website (www.toybox-study.eu) and was previously shown to be a reliable tool( Reference Gonzalez-Gil, Mouratidou and Cardon 22 ).
Descriptive statistics describing the baseline characteristics of the participating children per intervention or control group were conducted in the statistical software package IBM SPSS Statistics version 23.0. Differences between these two groups were tested using independent t tests for continuous variables or χ 2 tests for dichotomous variables. Missing data were imputed for the food items under study using the ‘multiple imputation’ macro in IBM SPSS Statistics version 23.0.
Mediation effects were assessed with the bootstrapping procedure of MacKinnon et al.( Reference MacKinnon, Lockwood and Williams 23 ) via the following mediation analysis steps: (i) estimating the effect of the intervention on unhealthy snacking and F&V consumption (τ coefficient); (ii) estimating the effect of the intervention on the proposed mediator (α coefficient); (iii) estimating the effect of the mediator on unhealthy snacking and F&V consumption (β coefficient) adjusting for the intervention effect (τ′ coefficient); (iv) computing the indirect effect of the intervention on unhealthy snacking and F&V consumption via the proposed mediator (αβ coefficient); and (v) bootstrapping the sampling distribution of αβ and deriving a bias-corrected CI with 5000 bootstrapped sampling distribution. All models were adjusted for age, sex, maternal education, country, baseline level of unhealthy snacking and F&V consumption, and baseline level of the mediator. STDY standardization methods (which mean that standardization was based only on the dependent variable, and not on both the dependent and independent variable because the independent variable (research group) was a binary variable) were applied. Adding a random intercept for kindergarten site did not improve the model fit and was therefore eliminated. Mediation analyses were run in Mplus version 8.0( Reference Androutsos, Katsarou and Payr 20 ).
Table 1 shows the baseline characteristics of the current study sample. Participating children were on average 4·75 years old, 49 % were female and 61 % had a mother with more than 14 years of education. Intervention and control group participants did not differ in terms of children’s age, sex, parental age, unhealthy snack consumption, F&V consumption or any of the baseline values of the potential mediators, with the exception of intervention children having less educated mothers, more parental self-efficacy with a nagging child and higher parental F&V consumption than control participants at baseline (P=0·03).
F&V, fruit(s) and vegetable(s).
*Significant difference between intervention and control groups based on independent t test for continuous variables or χ 2 test for dichotomous variables: P<0·05.
No significant total intervention effects were found on unhealthy snack consumption (c (unstandardized)=−2·41; 95 % CI −5·40, 0·39 (data not shown for the unstandardized analysis); τ (standardized)=−0·05; 95 % CI −0·12, 0·01) or F&V consumption (c=−1·59; 95 % CI −8·33, 5·00; τ=−0·01; 95 % CI −0·06; 0·04). As shown in Table 2, the intervention was effective in improving parental rules on snacking while watching television and snacking recommendation (i.e. permission of unhealthy snacking only during special occasions), parental unhealthy snack consumption, parental knowledge on children’s snacking recommendation and child’s attitude towards F&V as reported by their parent. Several potential mediators were found to be related to changes in unhealthy snacking and F&V consumption. Changes in parental rules (i.e. restriction) on unhealthy snacking while watching television and in permission of unhealthy snacking only during special occasions, improvement of parental unhealthy snack consumption and improvement of parental knowledge on snacking recommendation mediated the intervention effect on snacking consumption, whereas improvement in child’s attitude towards F&V as reported by their parent mediated the intervention effect on F&V consumption. Regarding interpretation of the results presented in Table 2, an example would be that an increase of 1 sd in the determinant ‘If I prohibit my child to eat sweet/salty snack I find it difficult to stick to my rules if (s)he starts nagging’ is associated with a decrease in unhealthy snack consumption of 0·07 sd, after adjusting for the intervention effect.
α, estimation of the effect of the intervention on the proposed mediator; β, estimation of the effect of the mediator on unhealthy snacking and F&V consumption; αβ, computation of the indirect effect of the intervention on unhealthy snacking and F&V consumption via the proposed mediator; τ′, estimation of the effect of the mediator on unhealthy snacking and F&V consumption adjusting for the intervention effect; F&V, fruit(s) and vegetable(s).
All models were single mediation models adjusted for child’s sex and age, maternal education, country, baseline values of snacking consumption and baseline values of the specific mediators. Bias-corrected bootstrapping using 5000 samples was conducted with the maximum likelihood estimator (except for categorical mediators indicated with †, which were conducted with the WLSMV indicator) using Mplus version 8.0. Standardized coefficients are shown using STDY standardization because of a binary independent variable. Significant associations are presented in bold font. The regression coefficients are presented in sd.
The ToyBox-intervention was a kindergarten-based, family-involved intervention aiming to prevent obesity at pre-school age via the promotion of healthy energy balance-related behaviours. The aim of the present study was to examine if the family-related determinants of snack consumption, which were identified and targeted in the ToyBox-study, mediated the effects of the ToyBox-intervention on pre-schoolers’ consumption of healthy and unhealthy snacks.
The present study showed that several family-related determinants of snack consumption mediated the effects of the ToyBox-intervention on pre-schoolers’ snack consumption. These mediators were parental rules (i.e. restriction) on unhealthy snacking while watching television, parental permission of unhealthy snacking only during special occasions, parental unhealthy snack consumption, parental knowledge on snacking recommendation and child’s attitude towards F&V as reported by their parent. Our findings are in line with previous studies. More specifically, a recent review showed that food availability and parental offering of foods have been successfully modified by several interventions and linked to positive changes in child outcomes( Reference Bekelman, Bellows and Johnson 24 ). Furthermore, a systematic review conducted by van Stralen et al. found evidence for attitude, knowledge and habit strength as mediators of interventions targeting dietary behaviour( Reference van Stralen, Yildirim and te Velde 25 ). In the case of the ToyBox-study, these determinants referred to the parents and teachers of the pre-school children, due to the children’s young age. Moreover, there is accumulated evidence highlighting that parental role modelling, which in the current study is depicted as limited parental F&V consumption, exerts a significant role in children’s F&V consumption( Reference Couch, Glanz and Zhou 26 ).
The ToyBox-intervention had a positive effect on several family-related determinants regarding children’s eating and snacking behaviour. Despite the strategies used to target these determinants being correctly and appropriately used, the ToyBox-intervention did not manage to significantly change children’s snack consumption. This observation is in line with similar intervention studies( Reference Freeman and Oliver 27 ). These findings could be attributed to the fact that dietary behaviour is one of the most complex behaviours and there may be more family-, school- or peer-related determinants than those identified during the preliminary phase of the ToyBox-study (focus groups as well as systematic literature reviews) that might also have an important role in determining dietary choices. Moreover, the results of the process evaluation conducted within the ToyBox-intervention showed that many parents did not read the provided newsletters and the tip-cards, which might explain the non-significant intervention effects regarding pre-schoolers’ snack food consumption (report submitted to the European Commission)( Reference Bekelman, Bellows and Johnson 24 , Reference Pinket, De Craemer and Maes 28 ). Furthermore, the duration of the intervention might not be adequate to change children’s snacking behaviour. Last but not least, like most school-based interventions, the ToyBox-intervention did not follow a personalized approach, thus was not tailored to individual needs and preferences that might further enhance its effectiveness.
Future intervention developers targeting pre-schoolers’ snack consumption are recommended to target the determinants that were found to mediate the intervention effect as well as the intervention strategies used in the ToyBox-intervention to change them, since those strategies succeeded in improving the targeted determinants. However, more extensive research should be done in the field of pre-schoolers’ snack food choices to capture the complete panel of determinants influencing pre-schoolers’ snacking behaviour. Furthermore, investing more time and effort to follow a more personalized approach in the population subgroups that are at higher risk (e.g. in families with low socio-economic status or in families with medical history of non-communicable diseases) might be more effective. Developing strategies to increase participants’ fidelity to the programme, such as making the material less burdensome or providing it via a web-based system, should also be considered in future public health initiatives.
Limitations and strengths
The present study has some strengths and limitations. The strengths include the large study sample, the diversity of countries, regions and socio-economic groups included, the standardization of all study procedures and tools, as well as the assessment of all determinants found to be associated with pre-school children’s snack consumption in the preliminary phases of the ToyBox-study. Furthermore, all measurements at pre- and post-test were taken during the same time period (i.e. May/June 2012 and 2013), thus limiting the potential seasonality effects on pre-schoolers’ snack consumption. On the other hand, data were self-reported by the parents, resulting in potential recall bias or social desirability of the responses. However, the reliability of the used questionnaires has been previously tested, before the start of the intervention, and they were found to be reliable tools( Reference Gonzalez-Gil, Mouratidou and Cardon 22 ). Furthermore, the data were collected in 2012–2013; however, this is to be expected from multicentre studies, which include complex procedures for data processing, handling and analysis. Even though attention on the health-related behaviours in young children is growing, the literature focusing on the pathways through which interventions such as the ToyBox-study improve health-related behaviours via family-related or other determinants is still scarce. Thus, the current study is strongly related and relevant to current policy and practice in order to improve the effectiveness of childhood obesity prevention programmes.
Even though the total effect of the ToyBox-intervention on pre-school children’s healthy and unhealthy snacking was not significant, the intervention was effective in improving parental rules on children’s unhealthy snack consumption (i.e. restriction while watching television and permission only at certain occasions) and parental consumption of unhealthy snacks, while it increased parental knowledge on snacking recommendations. Regarding the consumption of healthy snacks, the ToyBox-intervention improved children’s attitude towards F&V. All previously mentioned family-related determinants mediated the intervention effect on pre-schoolers’ consumption of healthy and unhealthy snacks. These findings imply that future interventions aiming to promote F&V consumption and limit the consumption of unhealthy snacks in pre-schoolers should target these mediators, but also target more determinants and use more strategies, such as personalization, to enhance their effectiveness.
Acknowledgements: The authors would like to thank the members of the ToyBox-study group. Financial support: This work was supported by the Seventh Framework Programme (CORDIS FP7) of the European Commission (grant number 245200). The European Commission had no role in the design, analysis or writing of this article. The content of this article reflects only the authors’ views and the European Community is not liable for any use that may be made of the information contained herein. The current study is part of a PhD that has been financially supported by the General Secretariat for Research and Technology (GSRT) and the Hellenic Foundation for Research and Innovation (HFRI) (number 133231/I2; 04/08/2017). GSRT and HFRI had no role in the design, analysis and writing of this article. Confict of interest: None. Authorship: All authors participated in the ToyBox-study and in the study design. All authors read, critically reviewed the manuscript and approved the final manuscript. M.M.v.S. analysed the data. C.-P.L. wrote the manuscript. Ethics of human subjects participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by ethics committees in the six participating countries (i.e. Ethical Committee of Ghent University Hospital (Belgium); Committee for the Ethics of Scientific Studies (KENI) at the Medical University of Varna (Bulgaria); Ethikkommission der Ludwig-Maximilians-Universität München (Germany); Ethics Committee of Harokopio University of Athens (Greece); Ethical Committee of Children’s Memorial Health Institute (Poland); and Comité Etico de Investigacion Clinica de Aragon (Spain)), in line with national regulations. Written informed consent was obtained from all subjects’ parents.
The ToyBox-study group consists of the following. Coordinator: Yannis Manios; Project Manager: Odysseas Androutsos; Steering Committee: Yannis Manios, Berthold Koletzko, Ilse De Bourdeaudhuij, Mai Chin A Paw, Luis Moreno, Carolyn Summerbell, Tim Lobstein, Lieven Annemans, Goof Buijs; External Advisors: John Reilly, Boyd Swinburn, Dianne Ward; Harokopio University (Greece): Yannis Manios, Odysseas Androutsos, Eva Grammatikaki, Christina Katsarou, Eftychia Apostolidou, Anastasia Livaniou, Katerina Lymperopoulou, Eirini Efstathopoulou, Christina-Paulina Lambrinou, Angeliki Giannopoulou, Evita Siatitsa, Efstathoula Argiri, Konstantina Maragkopoulou, Athanasios Douligeris; Ludwig-Maximilians-Universität München (Germany): Berthold Koletzko, Kristin Duvinage, Sabine Ibrügger, Angelika Strauß, Birgit Herbert, Julia Birnbaum, Annette Payr, Christine Geyer; Ghent University (Belgium), Department of Movement and Sport Sciences: Ilse De Bourdeaudhuij, Greet Cardon, Marieke De Craemer, Ellen De Decker and Department of Public Health: Lieven Annemans, Stefaan De Henauw, Lea Maes, Carine Vereecken, Jo Van Assche, Lore Pil; VU University Medical Center, EMGO Institute for Health and Care Research (the Netherlands): Mai Chin A Paw, Saskia te Velde; University of Zaragoza (Spain): Luis Moreno, Theodora Mouratidou, Juan Fernandez, Maribel Mesana, Pilar De Miguel-Etayo, Esther M. González-Gil, Luis Gracia-Marco, Beatriz Oves; Oslo and Akershus University College of Applied Sciences (Norway): Agneta Yngve, Susanna Kugelberg, Christel Lynch, Annhild Mosdøl, Bente B. Nilsen; University of Durham (UK): Carolyn Summerbell, Helen Moore, Wayne Douthwaite, Catherine Nixon; State Institute of Early Childhood Research (Germany): Susanne Kreichauf, Andreas Wildgruber; Children’s Memorial Health Institute (Poland): Piotr Socha, Zbigniew Kulaga, Kamila Zych, Magdalena Góźdź, Beata Gurzkowska, Katarzyna Szott; Medical University of Varna (Bulgaria): Violeta Iotova, Mina Lateva, Natalya Usheva, Sonya Galcheva, Vanya Marinova, Zhaneta Radkova, Nevyana Feschieva; International Association for the Study of Obesity (UK): Tim Lobstein, Andrea Aikenhead; CBO B.V. (the Netherlands): Goof Buijs, Annemiek Dorgelo, Aviva Nethe, Jan Jansen; AOK-Verlag (Germany): Otto Gmeiner, Jutta Retterath, Julia Wildeis, Axel Günthersberger; Roehampton University (UK): Leigh Gibson; University of Luxembourg (Luxembourg): Claus Voegele.
To view supplementary material for this article, please visit https://doi.org/10.1017/S1368980018002653